Data Center Network Topologies: VL2 (Virtual Layer 2)

Transcription

1 Data Center Network Topologies: VL2 (Virtual Layer 2) Hakim Weatherspoon Assistant Professor, Dept of Computer cience C 5413: High Performance ystems and Networking eptember 26, 2014 lides used and adapted judiciously from CO-561, Advanced Computer Networks At Princeton University

2 Goals for Today VL2: a scalable and flexible data center network A. Greenberg, J. R. Hamilton, N. Jain,. Kandula, C. Kim, P. Lahiri, D. A. Maltz, P. Patel, and. engupta. ACM Computer Communication Review (CCR), August 2009, pages

3 Architecture of Data Center Networks (DCN)

4 Conventional DCN Problems CR CR AR AR AR AR I want more 1:80 1:5 1: I have spare ones, but tatic network assignment Fragmentation of resource Poor server to server connectivity Traffics affects each other Poor reliability and utilization

5 Objectives: Uniform high capacity: Maximum rate of server to server traffic flow should be limited only by capacity on network cards Assigning servers to service should be independent of network topology Performance isolation: Traffic of one service should not be affected by traffic of other services Layer-2 semantics: Easily assign any server to any service Configure server with whatever IP address the service expects VM keeps the same IP address even after migration

6 Measurements and Implications of DCN Data-Center traffic analysis: Traffic volume between servers to entering/leaving data center is 4:1 Demand for bandwidth between servers growing faster Network is the bottleneck of computation Flow distribution analysis: Majority of flows are small, biggest flow size is 100MB The distribution of internal flows is simpler and more uniform 50% times of 10 concurrent flows, 5% greater than 80 concurrent flows

7 Measurements and Implications of DCN Traffic matrix analysis: Poor summarizing of traffic patterns Instability of traffic patterns Failure characteristics: Pattern of networking equipment failures: 95% < 1min, 98% < 1hr, 99.6% < 1 day, 0.09% > 10 days No obvious way to eliminate all failures from the top of the hierarchy

8 Virtual Layer 2 witch (VL2) Design principle: Randomizing to cope with volatility: Using Valiant Load Balancing (VLB) to do destination independent traffic spreading across multiple intermediate nodes Building on proven networking technology: Using IP routing and forwarding technologies available in commodity switches eparating names from locators: Using directory system to maintain the mapping between names and locations Embracing end systems: A VL2 agent at each server

9 Virtual Layer 2 witch (VL2) CR 1. L2 semantics... CR AR AR AR AR 2. Uniform high 3. Performance capacity isolation... 10

10 VL2 Goals and olutions Objective Approach olution 1. Layer-2 semantics 2. Uniform high capacity between servers 3. Performance Isolation Employ flat addressing Guarantee bandwidth for hose-model traffic Enforce hose model using existing mechanisms only Name-location separation & resolution service Flow-based random traffic indirection (Valiant LB) TCP 11 Hose : each node has ingress/egress bandwidth constraints

11 Name/Location eparation Cope with host churns with very little overhead witches run link-state routing and maintain only switch-level topology Allows to use low-cost switches Protects network and hosts from host-state churn Obviates host and switch reconfiguration ToR 1... ToR 2... ToR 3... ToR 4 Directory ervice x ToR 2 y ToR 3 z ToR 43 ToR 3 ToR 34 y z payload payload x y, z z Lookup & Response 12 ervers use flat names

12 Clos Network Topology Offer huge aggr capacity & multi paths at modest cost Int D (# of 10G ports)... Max DC size (# of ervers) 48 11,520 Aggr ,080 K aggr switches with D ports , TOR ervers 20*(DK/4) ervers

13 Valiant Load Balancing: Indirection Cope with arbitrary TMs with very little overhead I ANY I ANY [ ECMP + IP Anycast ] Harness huge bisection bandwidth I ANY Obviate esoteric traffic engineering or optimization Ensure robustness to failures Work with switch mechanisms available today T 1 T 2 T 3 T 4 T 5 T 6 Links used for up paths Links used for down paths I ANY T 35 yz payload x Equal 1. Must Cost spread Multi traffic Path Forwarding y2. Must ensure dst z independence 14

14 VL2 Directory ystem D RM D RM RM Directory 2. Reply 1. Lookup 3. Replicate 2. Reply D RM ervers 2. et 4. Ack (6. Disseminate) 1. Update 5. Ack ervers Agent Lookup Agent Update

15 Uniform high capacity: Evaluation All-to-all data shuffle stress test: 75 servers, deliver 500MB Maximal achievable goodput is 62.3 VL2 network efficiency as 58.8/62.3 = 94%

16 Fairness: Evaluation 75 nodes Real data center workload Plot Jain s fairness index for traffics to intermediate switches Fairness Index Aggr1 Aggr2 Aggr Time (s)

17 Performance isolation: Two types of services: Evaluation ervice one: 18 servers do single TCP transfer all the time ervice two: 19 servers starts a 8GB transfer over TCP every 2 seconds ervice two: 19 servers burst short TCP connections

18 Evaluation Convergence after link failures 75 servers All-to-all data shuffle Disconnect links between intermediate and aggregation switches

19 Perspective tudied the traffic pattern in a production data center and find the traffic patterns Design, build and deploy every component of VL2 in an 80 server testbed Apply VLB to randomly spreading traffics over multiple flows Using flat address to split IP addresses and server names

20 Critique The extra servers are needed to support the VL2 directory system,: Brings more cost on devices Hard to be implemented for data centers with tens of thousands of servers. All links and switches are working all the times, not power efficient No evaluation of real time performance.

21 VL2 vs. EATTLE imilar virtual layer 2 abstraction Flat end-point addresses Indirection through intermediate node Enterprise networks (eattle) Hard to change hosts directory on the switches parse traffic patterns effectiveness of caching Predictable traffic patterns no emphasis on TE Data center networks (VL2) Easy to change hosts move functionality to hosts Dense traffic matrix reduce dependency on caching Unpredictable traffic patterns ECMP and VLB for TE 22

22 Other Data Center Architectures VL2: A calable and Flexible Data Center Network consolidate layer-2/layer-3 into a virtual layer 2 separating naming and addressing, also deal with dynamic load-balancing issues A calable, Commodity Data Center Network Architecture a new Fat-tree inter-connection structure (topology) to increases bi-section bandwidth needs new addressing, forwarding/routing Other Approaches: PortLand: A calable Fault-Tolerant Layer 2 Data Center Network Fabric BCube: A High-Performance, erver-centric Network Architecture for Modular Data Centers

23 Ongoing Research 24

24 Research Questions What topology to use in data centers? Reducing wiring complexity Achieving high bisection bandwidth Exploiting capabilities of optics and wireless Routing architecture? Flat layer-2 network vs. hybrid switch/router Flat vs. hierarchical addressing How to perform traffic engineering? Over-engineering vs. adapting to load erver selection, VM placement, or optimizing routing Virtualization of NICs, servers, switches, 25

25 Research Questions Rethinking TCP congestion control? Low propagation delay and high bandwidth Incast problem leading to bursty packet loss Division of labor for TE, access control, VM, hypervisor, ToR, and core switches/routers Reducing energy consumption Better load balancing vs. selective shutting down Wide-area traffic engineering electing the least-loaded or closest data center ecurity Preventing information leakage and attacks 26

26 Before Next time Project Progress Need to setup environment as soon as possible And meet with groups, TA, and professor Lab0b Getting tarted with Fractus Use Fractus instead of Red Cloud Red Cloud instances will be terminated and state lost Due Monday, ept 29 Required review and reading for Friday, eptember 26 The Click Modular Router, E. Kohler, R. Morris, B. Chen, and M. F. Kaashoek. ACM ymposium on Operating ystems Principles (OP), December 1999, pages Check piazza: Check website for updated schedule